Normally error categorization requires human to set categorization keys for each error. But we do categorization automatically, and to effectively group errors we want to use error message format as part of the key.
Why format:
In the initial approach (to test how it works) I used file + line number. And the only problem here is that an inserted or a deleted line in the top of a file shifts the rest of the code. And line number of Errorf-s is not the same anymore. It leads to new errors categories. Which is obviously a bad categorization. And the plan is to replace line number with error format: error format is stable to changes, does not depend on variable values and distinctive about which error occurred.
In other words, this PR is required to automatically create tasks on new errors.
Normally error categorization requires human to set categorization keys for each error. But we do categorization automatically, and to effectively group errors we want to use error message format as part of the key.
Why format: In the initial approach (to test how it works) I used file + line number. And the only problem here is that an inserted or a deleted line in the top of a file shifts the rest of the code. And line number of
Errorf-s is not the same anymore. It leads to new errors categories. Which is obviously a bad categorization. And the plan is to replace line number with error format: error format is stable to changes, does not depend on variable values and distinctive about which error occurred.In other words, this PR is required to automatically create tasks on new errors.